The interactive coding interview course is free to use, free to self-host, and open to contributions

I built Skilled.dev 5 years ago based on my experience sitting on both sides of the table while interviewing at FAANG companies and hiring dozens of engineers. My goal is to help people land jobs they love where they can find meaning and purpose.
Skilled.dev is now free and fully open source. That includes all 66 lessons, the 28 coding problems you can run in your browser, and the video walkthroughs, which are now hosted on YouTube.
The code behind the site is MIT-licensed at github.com/leveluphq/skilled, so you can run your own copy or help improve this one. You don’t need to run your own copy to study. The hosted site at skilled.dev is free too.
If you find it helpful, give the GitHub repo a star so others can more easily discover it.
Interviews have changed a lot since I started the course, so I’ll start there.
How interviews have changed with AI
AI changed both what interviews test and how they test it, and companies went two ways.
Some now hand you an AI. Canva expects candidates to use tools like Copilot, Cursor, and Claude in its technical interviews, and in that round it asks, “Can they identify and fix issues in AI-generated code?” Meta, Google, and Datadog have run AI-assisted rounds too, and interview platforms like CoderPad and HackerRank sell them to everyone else.
Others take the AI away. Tools appeared that feed candidates answers during remote interviews without showing up on screen recordings, and some companies moved rounds back in person. Google gave fundamentals as its reason. Sundar Pichai said Google would add “at least one round of in-person interviews for people just to make sure the fundamentals are there.”
Both kinds of rounds test fundamentals. One checks whether you can write the code without help, and the other checks whether you can tell when the AI’s code is wrong. I go through both, with examples, in Coding Interview Prep When Agents Write the Code.
What the future may hold for interviews
My guess is that both kinds of rounds stay. AI-assisted rounds will spread, because they match how engineers work now. Companies will keep at least one round without AI, because they need to know what you can do when the AI is wrong. More interviews will start from existing code, where you find the bug, add the feature, or make it faster. Algorithm questions will get harder to memorize, and what you’ve built on your own will count for more.
Why fundamentals still matter
Sean Goedecke wrote that “The most important skill in prompting is expertise in the domain you’re prompting for” (LLMs reward expertise). If you know data structures and algorithms, you can steer an AI toward a good solution and catch it when it’s wrong. If you don’t, you have to take what it gives you.
Here’s one example from the deep dive. The brute-force answer to two-sum checks every pair of numbers:
function twoSum(nums, target) {
for (let i = 0; i < nums.length; i++) {
for (let j = i + 1; j < nums.length; j++) {
if (nums[i] + nums[j] === target) return [i, j];
}
}
return null;
}It’s correct, and with 1,000 numbers it finishes in under 2 milliseconds, so it passes a small test. With 100,000 numbers it takes 3.4 seconds. The version with a hash map takes 7.5 milliseconds. An AI can write either one in a second, and Big O is how you know which one you’re looking at.
What Skilled.dev is
Skilled.dev teaches what I’ve seen work in real coding interviews. The course covers 12 topics: arrays, Big O, hash tables, strings, recursion, searching and sorting, linked lists, stacks and queues, trees, graphs, dynamic programming, and JavaScript for frontend and Node roles. The topics most likely to come up, and the ones later topics build on, come first.

Each topic has two parts. First, a lesson teaches the data structure or algorithm, and you implement it from scratch. Then you practice interview questions that use it. Each question starts with a prompt like the one you’d get in an interview, and you try it yourself first. If you get stuck, you can step through an explanation of how to think through the problem. When you have a solution, you run it against test cases in the browser, then compare it with the optimal solution and its time and space complexity.

The 28 coding problems run in ten languages: JavaScript, TypeScript, Python, Java, C++, C#, Go, Kotlin, Swift, and Rust. Most lessons also have a written walkthrough in all ten languages.

65 lessons have a video walkthrough with chapters and written notes, and you can watch them without an account.

Outside the course, there are guides and interactive tools for topics like graph and tree traversal, linked-list pointers, recursion, and dynamic programming. There are also study plans for 7, 30, or 90 days, and a prep path for each language.

If you have an interview coming up, pick the study plan that fits the time you have. Otherwise, start at the beginning of the course and go in order. The full course needs a free account, and the videos, guides, tools, and language walkthroughs work without one. Try every problem yourself before you open the explanation, on paper or a whiteboard if you can. Now that AI is part of the job, I’d add one step: once you’ve solved a problem, ask an AI to review your solution, and decide whether you agree with what it says.
Free to use, and free to self-host
The code is at github.com/leveluphq/skilled under the MIT license. It’s a full-stack TypeScript app: a Next.js site with the lessons written in MDX, Hasura and Postgres for the backend, and a code runner that runs each submission in a throwaway Docker container.

You need Node.js 24, Docker, and the Hasura CLI. From the repo root, the README’s Quick Start looks like this:
nvm use
npm ci
npm run setup:env
docker compose -f hasura/docker-compose.yml up -d
export HASURA_GRAPHQL_ENDPOINT=http://localhost:8080
export HASURA_GRAPHQL_ADMIN_SECRET=<the local admin secret in hasura/docker-compose.yml>
hasura --project hasura migrate apply --database-name default
hasura --project hasura metadata apply
node scripts/course-seed.js apply
npm run runner:images javascript python
npm run dev
Then open http://localhost:3000/course. On your own copy, every lesson opens without an account, and Run Code runs in Docker on your machine. If a step fails, npm run check:setup tells you how to fix the first one that isn't done. You can also hand the setup to a coding agent, since the steps it needs are in the repo.
To run a public copy for a class, a bootcamp, or your team, you can run the code runner as a separate service on Cloudflare Containers or any Docker host. The runner docs cover both.
Contributions welcome
These are the areas where contributions help most:
- More lessons and explanations.
- More languages for the coding problems’ solutions and tests. Adding a Language explains how the runner works and what a new language needs.
- Easier local setup and better tooling.
Before you open a pull request, run the type check and the tests:
npm run ts
npm test
If you changed a lesson’s code or the runner, also run the conformance tests for what you changed. The conformance suite runs each lesson’s solutions and starting code in every language, and CI runs all of these checks on every pull request.
Doubling down on open source
The future is trending toward being completely open. When AI can clone a product effortlessly, there’s no reason tools and apps should be closed.
I plan to make all my tools open source. Some of the ones I’m most excited about are:
- Ri — An A-IDE (agentic integrated development environment) that works with your existing subscriptions. It combines context, task management, and agent orchestration to make productivity effortless.
- Beamd — An open-source proxy similar to Ngrok, but built for agents. It works completely from the CLI, no need to go to a web dashboard to register URLs.
- Agentex — Primitives for working with local and agent your subscriptions to providers like Codex and Claude Code.
Skilled.dev Is Now Free and Open Source was originally published in Level Up Coding on Medium, where people are continuing the conversation by highlighting and responding to this story.






